At Outlier.ai, I evaluate AI-generated responses using structured, user-interview-based assessment methods. On the Aether S2S Quality Screening project, I rank audio and text responses, detect hallucinations, and write detailed justifications.
With Handshake, I authored and evaluated coding and agentic task benchmarks against the Terminal-Bench standard for Project Dynamo. For Project Seal India, I crafted adversarial and edge-case prompts to expose failures in LLM web search and multi-step reasoning.
I built CodeForensics, an AST-based tool for detecting code plagiarism and AI-generated content, with a FastAPI backend and React frontend. I also developed an NLP resume-screening pipeline that extracts skills and matches resumes to job descriptions.
On the CampusPayX hackathon project, I led frontend development of role-based dashboards and QR-based payments. My other projects include a RAG-based PDF chatbot and a BERT-based fake-news classifier.
